A Machine Learning Algorithm Improves the Diagnostic Accuracy of the Histologic Component of Antibody Mediated Rejection (AMR-H) in Cardiac Transplant Endomyocardial Biopsies.

BACKGROUND: Pathologic antibody mediated rejection (pAMR) remains a major driver of graft failure in cardiac transplant patients. The endomyocardial biopsy remains the primary diagnostic tool but presents with challenges, particularly in distinguishing the histologic component (pAMR-H) defined by 1) intravascular macrophage accumulation in capillaries and 2) activated endothelial cells that expand the cytoplasm to narrow or occlude the vascular lumen. Frequently, pAMR-H is difficult to distinguish from acute cellular rejection (ACR) and healing injury. With the advent of digital slide scanning and advances in machine deep learning, artificial intelligence technology is widely under investigation in the areas of oncologic pathology, but in its infancy in transplant pathology. For the first time, we determined if a machine learning algorithm could distinguish pAMR-H from normal myocardium, healing injury and ACR. MATERIALS AND METHODS: A total of 4,212 annotations (1,053 regions of normal, 1,053 pAMR-H, 1,053 healing injury and 1,053 ACR) were completed from 300 hematoxylin and eosin slides scanned using a Leica Aperio GT450 digital whole slide scanner at 40X magnification. All regions of pAMR-H were annotated from patients confirmed with a previous diagnosis of pAMR2 (>50% positive C4d immunofluorescence and/or >10% CD68 positive intravascular macrophages). Annotations were imported into a Python 3.7 development environment using the OpenSlide™ package and a convolutional neural network approach utilizing transfer learning was performed. RESULTS: The machine learning algorithm showed 98% overall validation accuracy and pAMR-H was correctly distinguished from specific categories with the following accuracies: normal myocardium (99.2%), healing injury (99.5%) and ACR (99.5%). CONCLUSION: Our novel deep learning algorithm can reach acceptable, and possibly surpass, performance of current diagnostic standards of identifying pAMR-H. Such a tool may serve as an adjunct diagnostic aid for improving the pathologist's accuracy and reproducibility, especially in difficult cases with high inter-observer variability. This is one of the first studies that provides evidence that an artificial intelligence machine learning algorithm can be trained and validated to diagnose pAMR-H in cardiac transplant patients. Ongoing studies include multi-institutional verification testing to ensure generalizability.

Patient values and preferences regarding communicating risk versus benefit of benzodiazepine initiation: A cross-sectional survey study.

Background and Aims: Communicating information about the risks and benefits of benzodiazepines so that it is meaningful to the patient has not been previously described. This study aims to determine patient preferences regarding information received before initiating a benzodiazepine. Methods: An online survey was distributed through social media and advertisements to Canadians ≥18 years old over a 6-month period (May-Oct 2022) to collect participant's rating of importance of statements and factors about the risk and benefits of benzodiazepines before initiating treatment using a 10-point Likert-type scale. Treatment preferences based on efficacy and risk information were also elicited. The survey was developed and pilot-tested in collaboration with an advisory committee of individuals with lived and living experience with benzodiazepine use. Results: Thirty-seven participants responded to the survey (mean age 30 years old, 81.1% identified as female). The majority of respondents had a history of anxiety (83.8%) or insomnia (32.4%), and 10 (27.0%) respondents had used a benzodiazepine. Patient counseling related to withdrawal symptoms of benzodiazepines, risk of harm in combination with other sedating agents, risk of physical and psychological dependence, and risk of effects on cognition were rated high in the importance of receiving this information before starting a benzodiazepine relative to efficacy endpoints, such as improvement in sleep parameters. When provided with information about the chance of efficacy and risk of harm, 100% would have selected cognitive behavioral therapy as the best treatment option. The most frequently reported source of medication information where patients have sought information was from the internet (25.0%), followed by doctors (21.9%) and pharmacists (18.8%). Conclusions: This study identified patient important factors and statements viewed as important to communicate before initiating a benzodiazepine. The findings of this survey study will help inform decision-making when considering treatment options for managing anxiety or insomnia.

The Role of Machine Learning in Cardiovascular Pathology.

Machine learning has seen slow but steady uptake in diagnostic pathology over the past decade to assess digital whole-slide images. Machine learning tools have incredible potential to standardise, and likely even improve, histopathologic diagnoses, but they are not yet widely used in clinical practice. We describe the principles of these tools and technologies and some successful preclinical and pretranslational efforts in cardiovascular pathology, as well as a roadmap for moving forward. In nonhuman animal models, one proof-of-principle application is in rodent progressive cardiomyopathy, which is of particular significance to drug toxicity studies. Basic science successes include screening the quality of differentiated stem cells and characterising cardiomyocyte developmental stages, with potential applications for research and toxicology/drug safety screening using derived or native human pluripotent stem cells differentiated into cardiomyocytes. Translational studies of particular note include those with success in diagnosing the various forms of heart allograft rejection. For fully realising the value of these tools in clinical cardiovascular pathology, we identify 3 essential challenges. First is image quality standardisation to ensure that algorithms can be developed and implemented on robust, consistent data. The second is consensus diagnosis; experts don't always agree, and thus "truth" may be difficult to establish, but the algorithms themselves may provide a solution. The third is the need for large-enough data sets to facilitate robust algorithm development, necessitating large cross-institutional shared image databases. The power of histopathology-based machine learning technologies is tremendous, and we outline the next steps needed to capitalise on this power.

Frozen-state storage stability of a monoclonal antibody: aggregation is impacted by freezing rate and solute distribution.

Freezing of protein solutions perturbs protein conformation, potentially leading to aggregate formation during long-term storage in the frozen state. Macroscopic protein concentration profiles in small cylindrical vessels were determined for a monoclonal antibody frozen in a trehalose-based formulation for various freezing protocols. Slow cooling rates led to concentration differences between outer edges of the tank and the center, up to twice the initial concentration. Fast cooling rates resulted in much smaller differences in protein distribution, likely due to the formation of dendritic ice, which traps solutes in micropockets, limiting their transport by convection and diffusion. Analysis of protein stability after more than 6 months storage at either -10°C or -20°C [above glass transition temperature (T'g )] or -80°C (below T'g ) revealed that aggregation correlated with the cooling rate. Slow-cooled vessels stored above T'g exhibited increased aggregation with time. In contrast, fast-cooled vessels and those stored below T'g showed small to no increase in aggregation at any position. Rapid entrapment of protein in a solute matrix by fast freezing results in improved stability even when stored above T'g . © 2013 Wiley Periodicals, Inc. and the American Pharmacists Association J Pharm Sci 102:1194-1208, 2013.

Immunization with the SDPM1 peptide lowers amyloid plaque burden and improves cognitive function in the APPswePSEN1(A246E) transgenic mouse model of Alzheimer's disease.

Vaccination has become an important therapeutic approach to the treatment of Alzheimer's disease (AD), however, immunization with Abeta amyloid can have unwanted, potentially lethal, side effects. Here we demonstrate an alternative peptide-mimotope vaccine strategy using the SDPM1 peptide. SDPM1 is a 20 amino acid peptide bounded by cysteines that binds tetramer forms of Abeta(1-40)- and Abeta(1-42)-amyloids and blocks subsequent Abeta amyloid aggregation. Immunization of mice with SDPM1 induced peptide-mimotope antibodies with the same biological activity as the SDPM1 peptide. When done prior to the onset of amyloid plaque formation, SDPM1 vaccination of APPswePSEN1(A246E) transgenic mice reduced amyloid plaque burden and Abeta(1-40) and Abeta(1-42) levels in the brain, improved cognitive performance in Morris water maze tests, and resulted in no increased T cell responses to immunogenic or Abeta peptides or brain inflammation. When done after plaque burden was already significant, SDPM1 immunization still significantly reduced amyloid plaque burden and Abeta(1-40/1-42) peptide levels in APPswePSEN1(A246E) brain without inducing encephalitogenic T cell responses or brain inflammation, but treatment at this stage did not improve cognitive function. These experiments demonstrate the efficacy of a novel vaccine approach for Alzheimer's disease where immunization with an Abeta(1-40/1-42) amyloid-specific binding and blocking peptide is used to inhibit the development of neuropathology and cognitive dysfunction.

The synaptic CT carbohydrate modulates binding and expression of extracellular matrix proteins in skeletal muscle: Partial dependence on utrophin.

The CT carbohydrate, Neu5Ac/Neu5Gcalpha2,3[GalNAcbeta1,4]Galbeta1,4GlcNAcbeta-, is specifically expressed at the neuromuscular junction in skeletal myofibers of adult vertebrates. When Galgt2, the glycosyltransferase that creates the synaptic beta1,4GalNAc portion of this glycan, is overexpressed in extrasynaptic regions of the myofiber membrane, alpha dystroglycan becomes glycosylated with the CT carbohydrate and this coincides with the ectopic expression of synaptic dystroglycan-binding proteins, including laminin alpha4, laminin alpha5, and utrophin. Here we show that both synaptic and extrasynaptic forms of laminin and agrin have increased binding to the CT carbohydrate compared to sialyl-N-acetyllactosamine, its extrasynaptically expressed precursor. Muscle laminins also show increased binding to CT-glycosylated muscle alpha dystroglycan relative to its non-CT-containing glycoforms. Overexpression of Galgt2 in transgenic mouse skeletal muscle increased the mRNA expression of extracellular matrix (ECM) genes, including agrin and laminin alpha5, as well as utrophin, integrin alpha7, and neuregulin. Increased expression of ECM proteins in Galgt2 transgenic skeletal muscles was partially dependent on utrophin, but utrophin was not required for Galgt2-induced changes in muscle growth or neuromuscular development. These experiments demonstrate that overexpression of a synaptic carbohydrate can increase both ECM binding to alpha dystroglycan and ECM expression in skeletal muscle, and they suggest a mechanism by which Galgt2 overexpression may inhibit muscular dystrophy and affect neuromuscular development.

O-fucosylation of muscle agrin determines its ability to cluster acetylcholine receptors.

Protein O-fucosyltransferase 1 (Pofut1) transfers fucose to serine or threonine on proteins, including Notch receptors, that contain EGF repeats with a particular consensus sequence. Here we demonstrate that agrin is O-fucosylated in a Pofut1-dependent manner, and that this glycosylation can regulate agrin function. Fucosylation of recombinant C45 agrin, both active (neural, z8) and inactive (muscle, z0) splice forms, was eliminated when agrin was overexpressed in Pofut1-deficient cells or by mutation of a consensus site for Pofut1 fucosylation (serine 1726 in the EGF4 domain). Loss of O-fucosylation caused a gain of function for muscle agrin such that it stimulated AChR clustering and MuSK phosphorylation in cultured myotubes at levels normally only found with the neural splice form. Deletion of Pofut1 in cultured primary myotubes and in adult skeletal muscle increased AChR aggregation. In addition, Pofut1 gene and protein expression and Pofut1 activity of the EGF4 domain of agrin were modulated during neuromuscular development. These data are consistent with a role for Pofut1 in AChR aggregation during synaptogenesis via the regulation of the synaptogenic activity of muscle agrin.

Altered expression of natively glycosylated alpha dystroglycan in pediatric solid tumors.

Altered glycosylation and/or expression of dystroglycan have been reported in forms of congenital muscular dystrophy as well as in cancers of the breast, colon, and oral epithelium. To date, however, there has been no study of the expression of dystroglycan in pediatric solid tumors. Using a combination of immunostaining on tissue microarrays and immunoblotting of snap-frozen unfixed tissues, we demonstrate a significant reduction in native alpha dystroglycan expression in pediatric alveolar rhabdomyosarcoma (RMS), embryonal RMS, neuroblastoma (NBL), and medulloblastoma, whereas expression of beta dystroglycan, which is cotranslated with alpha dystroglycan, is largely unchanged. Loss of native alpha dystroglycan expression was significantly more pronounced in stage 4 NBL than in pooled samples of stage 1 and stage 2 NBL, suggesting that loss of native alpha dystroglycan expression increases with advancing tumor stage. Neuroblastoma and RMS samples with reduced expression of native alpha dystroglycan also showed reduced laminin binding in laminin overlay experiments. Expression of natively glycosylated alpha dystroglycan was not altered in several other pediatric tumor types when compared with appropriate normal tissue controls. These data provide the first evidence that alpha dystroglycan glycosylation and laminin binding to alpha dystroglycan are altered in certain pediatric solid tumors and suggest that aberrant dystroglycan glycosylation may contribute to tumor cell biology in patients with RMS, medulloblastoma, and NBL.